To protect in order to Serve, adverse effects of leniency programs in view of industry asymmetry by Daniel Leliefeld and Evgenia Motchenkova Free University of Amsterdam and TILEC March 2007
Motivation for the paper � Definition: "Leniency programs (LP)" grant total or partial immunity from fines to firms that collaborate with the authorities. � Empirical evidence: LPs improve welfare by increasing the number of detected cartels and by shortening the investigation. � Question of optimal design of LPs � How LPs work if there are asymmetries between the firms? � Is protection of leniency applicants necessary? 15 March 2007 2
Outline of the presentation � Motivation � Review of related literature and main innovations � Structure of the two-stage game � Solution of the game and derivation of SPNE � Legal and economic implications of the analysis � Conclusions and policy implications 15 March 2007 3
Motivation and policy implications � Possible counterproductive effects of LPs , when LPs are not properly designed Moderate LPs may greatly facilitate the enforcement of long term cartel agreements (Spagnolo (2004), Ellis and Wilson (2003)). When procedure of application for leniency is not confidential leniency may increase duration of cartel agreements, when penalties and rate of law enforcement are low (Motchenkova (2004)). � Problem of optimal design of LPs 1. Lenient or strict LPs ? 2. Whether the procedure of application for leniency should be confidential or open? 3. Timing of application for leniency. 4. Different treatment for the first and second reporter. � Current paper: issue of protection of leniency applicants 15 March 2007 4
Real Business Examples of Retaliation and Predatory Pricing � Health insurance sector in Netherlands during the period of privatization of health care insurance market. � An illustration of coercion through the threat of retaliation can be found in the leniency application of British Petrol (BP) in the Bitumen Cartel . During its existence the colluders managed to increase trust between its members through the design of a collective punishment strategy. � Recent price wars in the Dutch food retailers market 15 March 2007 5
Review of related literature � Motta and Polo (2003): Investigate the effects of LPs on the incentives of the firms to collude. � Spagnolo (2004): Concludes that only courageous LPs may completely and costless deter cartels. Mentions possible counterproductive effects of LPs. � Ellis and Wilson (2001) and Buccorossi and Spagnolo (2001): counterproductive effects of leniency programs � Bain (1949), Milgrom and Roberts (1984): predatory pricing and limit pricing � Scherer (1980): predatory pricing (the “long purse” story) 15 March 2007 6
Main innovations and policy implications � Ideally, properly designed LPs should induce firms to self-report (see Motta and Polo (2003) or Spagnolo (2004)); however, in this paper we show that, taking into account threat of retaliation (e.g. through predatory pricing), this expected effect of LPs may be reduced. � That's why the protection of leniency applicants from this kind of abuses by other members of cartel could be a good supplement to current leniency programs. This is the main message of the paper. 15 March 2007 7
Timing of the game � The game starts with the assumption that collusion can be sustained in equilibrium (i.e. the discount factor is large enough to sustain collusion). Hence, with this model we can only study the impact of LPs on the incentives of the firms to report in the cartels that have already been formed. � Stage 0: Antitrust authority announces parameters of the penalty scheme, p and F= απ , and parameters of the leniency program (assume, US system). � Stage 1: The small firm moves. It can choose between two actions: self-report or keep cartel secret (once and for all decision) � Stage 2: The big firm responds to the action of the small by choosing whether to punish the small firm (through predatory pricing) for reporting the cartel or to abstain from punishment (once and for all decision) 15 March 2007 8
Structure of the two-stage game Stage 1 Stage 2 ( 1 − β ) π m − S Predate (1) π δ απ δ p βπ + − − αβπ − m K m ( ) ( ) m m 1 − δ 1 − δ Report ( − 1 β ) π Not m Predate (2) βπ − αβπ m m ( 1 − β ) π m − S Not report Predate (3) π δ απ δ p βπ + − K − p αβπ − m m ( ) ( ) m m 1 − δ 1 − δ Not Predate ( ) ( ) 1 − β π α 1 − β π p (4) m − m ( ) ( ) 1 1 − δ − δ βπ αβπ p − m m ( ) ( ) 1 − δ 1 − δ 15 March 2007 9
Solution of the game (Stage 2) 1. Determination of threshold for collusion to be preferred strategy before leniency program is introduced � by comparing outcomes (3) and (4) If (3) > (4), then predatory pricing is more attractive than collusion for the bigger firm in both situations (with or without the availability of a leniency program). This happens when the discount factor is greater than the following threshold: , p , � � . K � � � � � � K � � m � 1 � �� 1 � K p � � � � Comparative static of the behavior of δ *(K,p, α ) with respect to the main parameters of the model shows that � � �� � K , p , � � � � �� � K , p , � � � � �� � K , p , � � � � �� � K , p , � � � 0 if p � � 1 or � 0 if p � � 1, � 0, � 0. � K � K � p � � 15 March 2007 10
Solution of the game (Stage 2) 2. Next, in order to ensure consistent behavior (meaning that collusion is sustainable and there are no incentives to predate in the absence of the possibility of self-reporting) we will � c � � � � � � consider only interval ,so that outcome (3) is ruled out and collusion is sustainable before the revelation game starts. 3. Determination of threshold for competitive pricing to be preferred over predatory pricing for bigger firm after application for leniency ((2)>(1)): K � � m � 1 � p � � � � � � K , p , � � . K � � � � � � K , p , � � � 0 if p � � 1 Comparative statics: � K This implies that when p α <1 (i.e. expected penalty is low) the equilibrium (2) is less likely to occur the smaller the size of K (i.e. the higher the asymmetry). Recall that K is the size of the buffer of Small, since it equals the cost of e.g. driving the smaller firm out of the market. 15 March 2007 11
Intuition This implies that when p α <1 (i.e. expected penalty is low) the equilibrium (2) is less likely to occur the smaller the size of K. Recall that K is the size of the buffer of Small, since it equals the cost of e.g. driving the smaller firm out of the market. After Small looses its buffer it can't sustain the losses associated with the predatory price setting. Intuitively this means that the greater the size difference (asymmetry), the lower K and therefore threshold δ { ∗ } will be lower when asymmetry is greater. It also implies that raising the risk of being fined will increase δ { ∗ }. Intuitively it means that the smaller the asymmetry and the higher the chance of a capture and substantial fine, the more likely the perceived discount rate is below the threshold δ { ∗ }. 15 March 2007 12
Solution of the game (Stage 1) 4. Next, we consider the decision of the smaller firm in the first stage given no predatory pricing is chosen by Big in the second stage of the game. (2) is preferred over (4) by Small (i.e. self-reporting is more attractive for Small) when the discount factor is lower than the following threshold: � � p � � � ��� � K , p , � � . This is a clear indication that raising the rate of capture and the proportional fine will make the smaller firm to choose equilibrium (2) over (4), and small will therefore decide to self-report instead of continuing to collude. 15 March 2007 13
Solution of the game (Stage 1) � Finally, we also have to compare the payoffs for Small in case outcome (1) arises and in case outcome (4) arises. Equilibrium (1) in the model is the situation in which strategies (report, predate) are employed by the smaller firm and the bigger firm respectively. Equilibrium (4) is collusive equilibrium. The payoff of equilibrium (4) is higher than the payoff in equilibrium (1) for Small (i.e. collusion is more attractive for the small firm) if discount factor is higher than the following threshold: � 1 � � � � m p � � S � � ���� � K , S , p , � � . � � � 1 � � � � m � S � ���� � � ��� when p � � 1 p � � 1 � 1 � � � � m � S . In this case we have also that � ���� � 1. � 1 � � � � m � S or when � ���� � � ��� when p � � 1 p � � 1 � 1 � � � � m � S . In this case we have also that � ���� � 1. � 1 � � � � m � S or when 15 March 2007 14
Recommend
More recommend
